Nintendo 64 RGB Blur fixing

[Woozle]

Side note: I finally got my hands on an UltraHDMI N64. I flashed the firmware to 1.05, turned deblur on with integer scaling, and started playing my favorites on the system... All I can say is HOLY SHIT! The picture quality was emulator-perfect (of course), and it honestly felt like I was experiencing for the first time what the N64 should have been like all along. An incredible mod with the perfect set of features. It should be the standard by which all HDMI mods are measured.

I agree, it really is a great mod kit. I had mine installed by Badass consoles in June.

Some minor complaints, it doesn't play nicely with my HDMI switch, the picture gets distorted. I also have some issues switching displays (unplug HDMI and plug it into another display while running), sometimes I don't get a picture and have to reset the console. Hi Def NES doesn't give me any of these issues.

I'm really liking these HDMI console kits (gcvideo, HDnes, ultrahdmi) since I no longer have to lug the framemeister and scart cables around when I travel.

Look forward to trying out the N64 VGA kit as well

[Josh128]

There is some crazy Japanese resistor mod that gets close to 1-Chip performance. Not sure how to find it again though.

http://vaot.mydns.jp/fc/sfc_sharp2.htm

Cool link, cool mod although WTF are they doing?!

This made me chuckle as it reminded me when many here busted my balls a couple years ago when I spoke of the horrible RGB out of the original SNES and how its S-Video out actually looked better.

[FBX]

This made me chuckle as it reminded me when many here busted my balls a couple years ago when I spoke of the horrible RGB out of the original SNES and how its S-Video out actually looked better.

Never encountered that scenario with the SNES. If the RGB was bad, so was the S-Video. At least on all the ones I've tested.

[Josh128]

^^

That is exactly what I was told back then, so I had to post pics, many of them! The colors and whites especially, bleed to the right with RGB output-- but on the same console on same TV using S-Video, the whites DO NOT bleed to the right. Technically why this happens, I cant say, but I assume it has something to do with where the RGB signal is picked up as it is converted to Y/C vs. where it is present on the multi out connector.

See many comparison pics and writeup below.



http://shmups.system11.org/viewtopic.php?f=6&t=53290

[FBX]

^^

That is exactly what I was told back then, so I had to post pics, many of them! The colors and whites especially, bleed to the right with RGB output-- but on the same console on same TV using S-Video, the whites DO NOT bleed to the right. Technically why this happens, I cant say, but I assume it has something to do with where the RGB signal is picked up as it is converted to Y/C vs. where it is present on the multi out connector.

Looks you used CRT screenshots. Have you tried direct capture from zoom-ins on a Framemeister? Reason I suggest this is the phenomenon could be the result of how the CRT processes the signals of RGB and S-Video. In the case of the Framemeister, you can zoom into the individual pixels and get a good look at whats going on (provided you use H_Scaler = 4 or 5 and V_Scaler = 6 or 7). RetroRGB and I did this when testing the PQ of various SNES consoles. You can see the pics here:

http://retrorgb.com/snesversioncompare.html

[Josh128]

^^

It occured on everything I tried it on (2 different CRTs and the F4500 plasma). I was able to check the plasma by using a GEFEN TV S-Video to HDMI box.

I didnt take pics of the S-Video on the plasma but I remember posting some RGB shots of FF3 on it somewhere around here. I dont have a direct capture device, but I dont see how that would do anything different-- are you saying it might show the SVideo bleed where it doesnt on other sets and devices?

[FBX]

Well it's certainly intriguing. I misplaced my S-Video cable, but I'll see if I can hunt one down and check it out on the Framemeister. It's quite possible the SNES somehow gets rid of the bleed when processing the picture into S-Video.

[mikejmoffitt]

Those pictures look like ringing. Turn down the "Sharpness" or aperture control, and the set may need a recap on the signal section.

It's all a little offtopic for this thread, though...

[copy]

I see the blurring on my non-1CHIP SNES through s-video to the Framemeister. I also found a YouTube video with the same setup which shows it:

SNES Gameplay scaled through XRGB-Mini Framemeister! 1080p with and without scanlines!

(Incidentally, I just received my first 1CHIP in the mail today, but haven't had time to try it yet.)

[Josh128]

I see the blurring on my non-1CHIP SNES through s-video to the Framemeister. I also found a YouTube video with the same setup which shows it:

SNES Gameplay scaled through XRGB-Mini Framemeister! 1080p with and without scanlines!

(Incidentally, I just received my first 1CHIP in the mail today, but haven't had time to try it yet.)

Copy, nothing in this youtube video shows the kind of color bleed shown in both my SNES 1 RGB shots and the shots from the Japanese resistor mod thread. It doesnt even compare the SVideo to the RGB, its strictly SVideo here. I just updated the WEGA thread with a new comparison on the F4500 plasma. Im not saying the SVideo is perfect, but its far superior to the RGB out, its what Ive said all along.

Mike, the white bleed has nothing to do with the CRT. Indeed there may some ringing in the SVideo pics as I dont remember what sharpness setting I used, but theres very little if any in the SNES 2 RGB shots. What little there is, is probably due to my sharpness setting preferences. Anyways youre right this is too O/T, so I'm out.

[FBX]

Im not saying the SVideo is perfect, but its far superior to the RGB out, its what Ive said all along.

If true, it would entirely depend on the console. My 1CHIP-03 and mini both have superb RGB PQ. But yeah, when I get my S-Video cables in the mail, I'll switch over to your thread.

[copy]

My mistake, I was conflating the color bleeding with the usual blurring issue. Sorry to jump in.

[andykara2003]

FBX - He's talking about the original (non 1-chip) SNES. Should be in another thread tho.

[FBX]

With regard to the UltraHDMI mod's deblur function, has anyone else noticed it introduces faint vertical jail bars that are about 1 pixel apart? I noticed this when looking at a large solid color on the screen. When I toggled the deblur function off, the jail bars vanished. When I toggled it back on, they appeared again. Then I thought maybe the jail bars were always there and deblur just 'reveals' them, but I was able to disprove this by loading up Quake II, which has no blur by default (in 8MB expansion mode). There again, turning the function on caused jail bars.

[Ikaruga11]

ruh roh

[bobrocks95]

With regard to the UltraHDMI mod's deblur function, has anyone else noticed it introduces faint vertical jail bars that are about 1 pixel apart? I noticed this when looking at a large solid color on the screen. When I toggled the deblur function off, the jail bars vanished. When I toggled it back on, they appeared again. Then I thought maybe the jail bars were always there and deblur just 'reveals' them, but I was able to disprove this by loading up Quake II, which has no blur by default (in 8MB expansion mode). There again, turning the function on caused jail bars.

Is that with all games or just some? Games that aren't using the blur already, like Quake II, might be showing the jailbars.

[FBX]

With regard to the UltraHDMI mod's deblur function, has anyone else noticed it introduces faint vertical jail bars that are about 1 pixel apart? I noticed this when looking at a large solid color on the screen. When I toggled the deblur function off, the jail bars vanished. When I toggled it back on, they appeared again. Then I thought maybe the jail bars were always there and deblur just 'reveals' them, but I was able to disprove this by loading up Quake II, which has no blur by default (in 8MB expansion mode). There again, turning the function on caused jail bars.

Is that with all games or just some? Games that aren't using the blur already, like Quake II, might be showing the jailbars.

All games. Quake II was just used to show that it was coming from the deblur function itself, since the game is already crystal clear by default (in 8MB mode). Also I noticed the jail bars in My Life in Gaming's video on the UltraHDMI mod.

[FBX]

I got a screengrab going below scaled to 2x from 5x integer so you can see a zoom in (left is deblur off, right is deblur on):



It happens in all games. It's one of those deals where once you see it, you can't help but notice it every time you turn vi deblur on.

Edit: Tested some more games. It's harder to spot in games with dithering because there's no solid colors view close up. For example in F-Zero, you can see the jail bars in the N64 logo at the beginning, but in-game racing has so much dithering going on that you can't really see them.

Edit 2: Also discovered "integer +" (5x in 1080p) isn't set to square pixels on the horizontal axis, whereas "integer" (4x in 1080p) is.

[telemetry]

Just wild mass speculation.

Problem with the deblur algorithm: the algorithm at a high level assumes that every other output pixel column is an average of the two surrounding ones. Perhaps a solid vertical color block is able to "fool" the algorithm into thinking it is an average of alternating dark & light pixels.

OR, even crazier thought:

Problem with the N64: solid color is misrendered with artifacts by the N64, *and* the blur actually helped hide that by averaging over the artifacts. Since the deblur is supposedly just "tossing" extra pixels, not changing the authentic ones, maybe the N64 native video buffer itself has output artifacts.

You could confirm that the artifacts aren't due to scaling — is there a way to force the UltraHDMI to output 1:1 or 240p over HDMI?

[FBX]

Just wild mass speculation.

Problem with the deblur algorithm: the algorithm at a high level assumes that every other output pixel column is an average of the two surrounding ones. Perhaps a solid vertical color block is able to "fool" the algorithm into thinking it is an average of alternating dark & light pixels.

OR, even crazier thought:

Problem with the N64: solid color is misrendered with artifacts by the N64, *and* the blur actually helped hide that by averaging over the artifacts. Since the deblur is supposedly just "tossing" extra pixels, not changing the authentic ones, maybe the N64 native video buffer itself has output artifacts.

It's difficult to say. Both Doom 64 and Quake II (in 8MB mode) do NOT require deblur for perfect pixels (they already are sharp by default), and since the jail bars ONLY show up when deblur is turned on in even in these games, I have to believe the bars are artifacts of the deblur function, and not inherent to the N64 itself.

[Ikaruga11]

So does that mean these jailbars will show up on an RGB modded N64 using borti's deblur?

[FBX]

So does that mean these jailbars will show up on an RGB modded N64 using borti's deblur?

No clue. I'd like to know what exactly causes it before we can even speculate on that.

[FBX]

Here's a screenshot in Mortal Kombat Trilogy in 4x integer mode 1080p. You can see the jail bars in the dark purple at the top, as well as in the light purple around the waist level of the fighters:

Spoiler

[leonk]

WOW. Would love to know if others noticed this with their HDMI N64 as well.

I've installed a few dozen N64RGB's with Borti's firmware .. no jail bars ever noticed. My PVM is really good at making jail bars pop; That or sitting 2' away from the screen will do it also.

[RGB32E]

Most Midway titles aren't 320x240, so de-blur artifacts are misguided.

INTEGER+
Multiplies each pixel by 1.5x (720p) or 2.5x (1080p). When a game uses 320x240 progressive mode, each pixel will be exactly 3x3 or 5x5 pixels. Produces visual artifacts on anything higher-res

If a given game has a gradient pattern of sorts, there can be scenarios where it "looks" like jail bars, but aren't due to any sort of actual interference pattern, so one shouldn't get the two confused...

Edit 2: Also discovered "integer +" (5x in 1080p) isn't set to square pixels on the horizontal axis, whereas "integer" (4x in 1080p) is.

Math is hard for you?

Here's a 5x5 checkerboard overlay. If it wasn't 5x5, you'd have a mismatch, which isn't the case. It's more that 320x240 mode doesn't work out to quite that many visible pixels, so dividing the visible picture width by 5 is a math fail boat.

[CaveManGamer]

Not sure if this has been discussed earlier, but I've recently had my N64 modded with Tim's board and Borti's deblur. I was playing around with it the other day and thought I'd try out my GS to remove AA blur. The dithering was AWFUL, but I realized that I could no longer toggle 15 bit colour mode on and off. I've now tried it on over 10 games and the results are always the same. Any ideas?
Just thought it odd that this feature is disabled when using a GS.

[Ikaruga11]

Not sure if this has been discussed earlier, but I've recently had my N64 modded with Tim's board and Borti's deblur. I was playing around with it the other day and thought I'd try out my GS to remove AA blur. The dithering was AWFUL, but I realized that I could no longer toggle 15 bit colour mode on and off. I've now tried it on over 10 games and the results are always the same. Any ideas?
Just thought it odd that this feature is disabled when using a GS.

That is strange. I don't see how using a GameShark could disable something so specific like that in borti's firmware.

The general consensus is that the removing the hardware-based blur (UltraHDMI, borti's firmware, etc.) on the Nintendo 64 works wonders and is far better than leaving it on.

On the other hand, removing the software-based blur in games (using GameShark) is pretty bad because it introduces a lot of visual quirks and heavy aliasing like you mentioned in your post.

My Life in Gaming has a very cohesive and thorough video on both types of blur removal and video comparisons between all four possible outputs: (normal, primary disabled, secondary disabled, primary + secondary disabled) https://www.youtube.com/watch?v=QDiHgKil8AQ

[FBX]

Most Midway titles aren't 320x240, so de-blur artifacts are misguided.

That was just one game example. Every game I've tried shows the same artifacts in certain shades of solid color. So maybe instead of "Most Midway titles", you should be stating "Most N64 titles", which therein makes the artifact issue more relevant. You also tried to use a gradient argument, when I already stated the artifacts show up in plain solid colors ONLY when the deblur function is turned on. The dead giveaway is Quake 2's pause menu. By default the game is already crystal clear and doesn't need deblur, and so the pause menu's blue tone is perfectly clean. When you turn deblur on, the jail bars show up and are easily spotted. Give it a try yourself (assuming you have access to an UltraHDMi N64 that is). But as I already stated, Quake 2 doesn't need deblur, so I thought it might be an issue related to that, however, since I see it in many other games that do need deblur, I ruled that out as the cause.

Math is hard for you?

Nope, just going by the edge artifacts in Doom 64. Try out your pattern on the menu screen with the red text. There are edges that are randomly blurred in 5x, indicating a non-integer scale. Does this game have some sort of non-standard resolution that causes this, and if so, why does it look perfectly scaled at 4x? Please do tell as I want to understand what's going on with that game.

Edit: And here's a 2x zoom in of "Integer" versus "integer+" on Doom 64:



Edit2: My theory is Doom 64 must be behaving as a high-res game, and technically isn't low res even though the pixels we see are. Maybe the same trick Quake II is doing?

[FBX]

Out of curiosity, does anyone (besides RGB32E) feel my question about the 4x versus 5x issue is foolish or stupid or lacking in reason? Just want to make sure I'm not sounding like a complete idiot to everyone, or if my question seems reasonable to ask. Because I get the sense from him that I shouldn't even be questioning these things about the display issues of these functions, because nobody else would be so foolish as to not understand what's going on.

[telemetry]

Having watched and read a few UltraHDMI reviews repeatedly (Phonedork, MLIG, retroRGB, videogameperfection) as well as perusing the UltraHDMI website, it's a little confusing how "Integer+" behaves differently from other modes.

I get the scaling (Integer> 2x/720p, Integer+> 3x/720p), which means in Integer+ the vertical pixels don't really align to an even multiple (but an odd multiple) of the display's vertical resolution. So it's understandable why scanlines wouldn't be consistent.

But your captures of the "blurry" scaling doesn't seem consistent with my understanding of the behavior (Integer+ should have vertical alignment issues for scanlines, but the horizontal still sounds like a clean 3x scale).

EDIT: replace 3x scaling with 5x for the 1080p Integer+ mode).